Add <none> placeholder to clear / unset inherited value

From a usability perspective I'd prefer something in the UI, e.g. a checkbox that lets you control whether the value is overridden. Let's pick this up in #1798984.

Reopening this as a simpler solution that won't involve greatly expanding the DOM structure and JS logic on edit forms.

So something like this might work, maybe?

Works great! Thanks for the fix @damien!

I basically copied the example by @anybody, and it works like a charm! The canonical url will get removed on 403 pages if we declare it as .

I'll add a test for this.

To be honest, I don't quite get this request and how the inheritance works for metatags?

Based on the request, it sounds like, if we would set the metatag value of a child metatag to an empty string (e.g. setting the 403 canonical_url to ""), the metatag would use the parents value (global's canonical url). But this isn't the case.

Setting an empty string instead of already does the exact same thing. This is also represented in code and the currently failing tests.

    // If there is no value, just return either an empty array or empty string.
    if (is_null($this->value) || $this->value == '') {
      return [];
    }

    // Support the "<none>" option to hide all output for this meta tag.
    if (!is_array($this->value) && $this->value == '<none>') {
      return [];
    }

And how does the whole inheritance work then?

403

Inherits meta tags from: Global

I don't think it does.

If we allowed to set an empty string, inherit the value from the parent and delete the metatag, @thomas.frobieter, and I suggest having the following changes:

• Add the ability to provide <none> as value, to kill the metatag entirely.
• Add the ability to provide <empty> as value, to let the metatag have an empty string.
• Add the ability to provide to leave the input empty to use the parent value (inherited).

Note, that the inherit / empty string functionality could also be implemented via checkbox.

All done, please review.

Yea, this whole endeavour was caused, because programmatically setting a metatag value to an empty string will kill the metatag output entirely, resulting in the same behaviour as providing <none> (which is the root cause of my confusion).

Only if we set an empty string via UI, the submit call will silently override our empty string with its parent value. That is also, why the tests failed originally. This is REALLY confusing! ESPECIALLY if there is also an empy / isNull check on the ouput handler and no way to tell, that $this->value will silently get overriden on submit...

Created the follow-up issue here: #3500307: Set metatag parent programmatically (inherited value).

What about values coming from <select> ?
We have some in the Twitter card, but mostly in Schema.org Metatag module that use a lot of them.
Is there a way to automatically add <none> to the <select> or this is due to this module ?

tostinni: That's a great point - please add a separate issue to extend tags to provide the "none" option. Thank you.

Committed. Thanks everyone, and Anybody ;-)